AI Voice Agent Cost Calculator
Estimate telephony, voice platform, speech-to-text, text-to-speech, and LLM costs for AI voice agents. Add margin to see what to charge per minute or per call.
Call volume
Voice stack
List pricing for US local numbers; toll-free and international rates differ.
Includes orchestration + telephony piping; usage tiers apply.
Low-latency streaming ASR with punctuation and diarization.
High-quality multilingual voices; Turbo tier is cheaper.
Fast, cheap voice-agent default.
Usage assumptions
Monthly minutes
—
Cost per minute
—
Cost per call
—
Total monthly cost
—
Telephony
—
Voice platform
—
STT + TTS
—
LLM
—
Pricing to hit margin
Price per minute
—
includes overhead
Price per call
—
Required monthly revenue
—
Cost by voice platform
Same call volume and stack, swapping only the voice platform.
| Platform | Platform/mo | Total/mo |
|---|---|---|
| Vapi $0.035/min | — | — |
| Retell AI $0.030/min | — | — |
| Bland AI $0.040/min | — | — |
| Synthflow $0.025/min | — | — |
| Pipecat (self-hosted) $0.000/min | — | — |
| Daily.co (custom) $0.004/min | — | — |
Cost by telephony provider
Same call volume, swapping only telephony. Assumes inbound minutes.
| Provider | Telephony/mo | Total/mo |
|---|---|---|
| Twilio $0.0085/min | — | — |
| Vonage (Nexmo) $0.0050/min | — | — |
| Plivo $0.0050/min | — | — |
| Bandwidth $0.0040/min | — | — |
How the estimate works
- Telephony = minutes × per-minute rate + phone numbers × monthly fee + SMS count × SMS rate.
- Voice platform = minutes × platform fee per minute.
- STT = minutes × speech-to-text per-minute rate.
- TTS = (minutes × chars per minute ÷ 1,000) × TTS per-1k-char rate.
- LLM = minutes × (input tokens/1k × input price + output tokens/1k × output price).
- Margin pricing = cost per minute × (1 + overhead%) ÷ (1 − margin%).
When to use each stack
- • Managed platforms (Vapi/Retell/Bland) for fastest time-to-market and built-in latency tuning.
- • Pipecat + Daily when you need deep customization and can self-host.
- • Deepgram for cheap, fast ASR; AssemblyAI when accuracy matters most.
- • ElevenLabs for premium voices; OpenAI / Amazon Polly for cost-sensitive scale.
- • GPT-4o mini / Claude Haiku for most voice agents; upgrade for complex reasoning.
Frequently asked questions
What does an AI voice agent cost per minute?▼
A typical managed stack costs $0.05–$0.15 per minute depending on telephony, speech-to-text, text-to-speech, LLM, and voice platform choices. Self-hosted stacks can drop variable cost close to $0 but add infrastructure labor.
Which voice platform is cheapest?▼
Open-source frameworks like Pipecat have no platform fee, but you pay for compute, telephony, and APIs. Among managed platforms, Synthflow and Retell often have the lowest per-minute platform fees, while Vapi and Bland include more bundled orchestration.
How much do AI voice agent telephony costs add?▼
US inbound local numbers typically cost $0.004–$0.009 per minute plus a $0.35–$1.15 monthly number fee. Outbound calls and SMS are billed separately; high-volume outbound use cases like reminders can make telephony the largest cost.
Are LLM costs significant in voice agents?▼
For cheap models like GPT-4o mini, LLM cost is usually under $0.01 per minute. For GPT-4o or Claude Sonnet it can become the second- or third-largest cost, especially on long calls with large context windows.
How do I price a voice-agent product with margin?▼
Use the cost-per-minute from this calculator, then divide by (1 − target margin). For example, if cost is $0.08/min and you want 30% gross margin, charge about $0.114/min. Add overhead and support costs on top for true net margin.
Prices are directional estimates based on public list pricing. Confirm current rates with each provider before committing to a pricing model. Last updated: 2026-08-07.